| blob | 74d9ee2e70bf0a247b47ddd2b426d4b19698d9dc |
1 Git 2.15 Release Notes
2 ======================
4 Backward compatibility notes and other notable changes.
6 * Use of an empty string as a pathspec element that is used for
7 'everything matches' is still warned and Git asks users to use a
8 more explicit '.' for that instead. The hope is that existing
9 users will not mind this change, and eventually the warning can be
10 turned into a hard error, upgrading the deprecation into removal of
11 this (mis)feature. That is now scheduled to happen in Git v2.16,
12 the next major release after this one.
14 * Git now avoids blindly falling back to ".git" when the setup
15 sequence said we are _not_ in Git repository. A corner case that
16 happens to work right now may be broken by a call to BUG().
17 We've tried hard to locate such cases and fixed them, but there
18 might still be cases that need to be addressed--bug reports are
19 greatly appreciated.
21 * "branch --set-upstream" that has been deprecated in Git 1.8 has
22 finally been retired.
25 Updates since v2.14
26 -------------------
28 UI, Workflows & Features
30 * An example that is now obsolete has been removed from a sample hook,
31 and an old example in it that added a sign-off manually has been
32 improved to use the interpret-trailers command.
34 * The advice message given when "git rebase" stops for conflicting
35 changes has been improved.
37 * The "rerere-train" script (in contrib/) learned the "--overwrite"
38 option to allow overwriting existing recorded resolutions.
40 * "git contacts" (in contrib/) now lists the address on the
41 "Reported-by:" trailer to its output, in addition to those on
42 S-o-b: and other trailers, to make it easier to notify (and thank)
43 the original bug reporter.
45 * "git rebase", especially when it is run by mistake and ends up
46 trying to replay many changes, spent long time in silence. The
47 command has been taught to show progress report when it spends
48 long time preparing these many changes to replay (which would give
49 the user a chance to abort with ^C).
51 * "git merge" learned a "--signoff" option to add the Signed-off-by:
52 trailer with the committer's name.
54 * "git diff" learned to optionally paint new lines that are the same
55 as deleted lines elsewhere differently from genuinely new lines.
57 * "git interpret-trailers" learned to take the trailer specifications
58 from the command line that overrides the configured values.
60 * "git interpret-trailers" has been taught a "--parse" and a few
61 other options to make it easier for scripts to grab existing
62 trailer lines from a commit log message.
64 * The "--format=%(trailers)" option "git log" and its friends take
65 learned to take the 'unfold' and 'only' modifiers to normalize its
66 output, e.g. "git log --format=%(trailers:only,unfold)".
68 * "gitweb" shows a link to visit the 'raw' contents of blbos in the
69 history overview page.
71 * "[gc] rerereResolved = 5.days" used to be invalid, as the variable
72 is defined to take an integer counting the number of days. It now
73 is allowed.
75 * The code to acquire a lock on a reference (e.g. while accepting a
76 push from a client) used to immediately fail when the reference is
77 already locked---now it waits for a very short while and retries,
78 which can make it succeed if the lock holder was holding it during
79 a read-only operation.
81 * "branch --set-upstream" that has been deprecated in Git 1.8 has
82 finally been retired.
84 * The codepath to call external process filter for smudge/clean
85 operation learned to show the progress meter.
87 * "git rev-parse" learned "--is-shallow-repository", that is to be
88 used in a way similar to existing "--is-bare-repository" and
89 friends.
91 * "git describe --match <pattern>" has been taught to play well with
92 the "--all" option.
94 * "git branch" learned "-c/-C" to create a new branch by copying an
95 existing one.
97 * Some commands (most notably "git status") makes an opportunistic
98 update when performing a read-only operation to help optimize later
99 operations in the same repository. The new "--no-optional-locks"
100 option can be passed to Git to disable them.
102 * "git for-each-ref --format=..." learned a new format element,
103 %(trailers), to show only the commit log trailer part of the log
104 message.
107 Performance, Internal Implementation, Development Support etc.
109 * Conversion from uchar[20] to struct object_id continues.
111 * Start using selected c99 constructs in small, stable and
112 essentialpart of the system to catch people who care about
113 older compilers that do not grok them.
115 * The filter-process interface learned to allow a process with long
116 latency give a "delayed" response.
118 * Many uses of comparision callback function the hashmap API uses
119 cast the callback function type when registering it to
120 hashmap_init(), which defeats the compile time type checking when
121 the callback interface changes (e.g. gaining more parameters).
122 The callback implementations have been updated to take "void *"
123 pointers and cast them to the type they expect instead.
125 * Because recent Git for Windows do come with a real msgfmt, the
126 build procedure for git-gui has been updated to use it instead of a
127 hand-rolled substitute.
129 * "git grep --recurse-submodules" has been reworked to give a more
130 consistent output across submodule boundary (and do its thing
131 without having to fork a separate process).
133 * A helper function to read a single whole line into strbuf
134 mistakenly triggered OOM error at EOF under certain conditions,
135 which has been fixed.
137 * The "ref-store" code reorganization continues.
139 * "git commit" used to discard the index and re-read from the filesystem
140 just in case the pre-commit hook has updated it in the middle; this
141 has been optimized out when we know we do not run the pre-commit hook.
142 (merge 680ee550d7 kw/commit-keep-index-when-pre-commit-is-not-run later to maint).
144 * Updates to the HTTP layer we made recently unconditionally used
145 features of libCurl without checking the existence of them, causing
146 compilation errors, which has been fixed. Also migrate the code to
147 check feature macros, not version numbers, to cope better with
148 libCurl that vendor ships with backported features.
150 * The API to start showing progress meter after a short delay has
151 been simplified.
152 (merge 8aade107dd jc/simplify-progress later to maint).
154 * Code clean-up to avoid mixing values read from the .gitmodules file
155 and values read from the .git/config file.
157 * We used to spend more than necessary cycles allocating and freeing
158 piece of memory while writing each index entry out. This has been
159 optimized.
161 * Platforms that ship with a separate sha1 with collision detection
162 library can link to it instead of using the copy we ship as part of
163 our source tree.
165 * Code around "notes" have been cleaned up.
166 (merge 3964281524 mh/notes-cleanup later to maint).
168 * The long-standing rule that an in-core lockfile instance, once it
169 is used, must not be freed, has been lifted and the lockfile and
170 tempfile APIs have been updated to reduce the chance of programming
171 errors.
173 * Our hashmap implementation in hashmap.[ch] is not thread-safe when
174 adding a new item needs to expand the hashtable by rehashing; add
175 an API to disable the automatic rehashing to work it around.
177 * Many of our programs consider that it is OK to release dynamic
178 storage that is used throughout the life of the program by simply
179 exiting, but this makes it harder to leak detection tools to avoid
180 reporting false positives. Plug many existing leaks and introduce
181 a mechanism for developers to mark that the region of memory
182 pointed by a pointer is not lost/leaking to help these tools.
184 * As "git commit" to conclude a conflicted "git merge" honors the
185 commit-msg hook, "git merge" that records a merge commit that
186 cleanly auto-merges should, but it didn't.
188 * The codepath for "git merge-recursive" has been cleaned up.
190 * Many leaks of strbuf have been fixed.
192 * "git imap-send" has our own implementation of the protocol and also
193 can use more recent libCurl with the imap protocol support. Update
194 the latter so that it can use the credential subsystem, and then
195 make it the default option to use, so that we can eventually
196 deprecate and remove the former.
198 * "make style" runs git-clang-format to help developers by pointing
199 out coding style issues.
201 * A test to demonstrate "git mv" failing to adjust nested submodules
202 has been added.
203 (merge c514167df2 hv/mv-nested-submodules-test later to maint).
205 * On Cygwin, "ulimit -s" does not report failure but it does not work
206 at all, which causes an unexpected success of some tests that
207 expect failures under a limited stack situation. This has been
208 fixed.
210 * Many codepaths have been updated to squelch -Wimplicit-fallthrough
211 warnings from Gcc 7 (which is a good code hygiene).
213 * Add a helper for DLL loading in anticipation for its need in a
214 future topic RSN.
216 * "git status --ignored", when noticing that a directory without any
217 tracked path is ignored, still enumerated all the ignored paths in
218 the directory, which is unnecessary. The codepath has been
219 optimized to avoid this overhead.
221 * The final batch to "git rebase -i" updates to move more code from
222 the shell script to C has been merged.
224 * Operations that do not touch (majority of) packed refs have been
225 optimized by making accesses to packed-refs file lazy; we no longer
226 pre-parse everything, and an access to a single ref in the
227 packed-refs does not touch majority of irrelevant refs, either.
229 * Add comment to clarify that the style file is meant to be used with
230 clang-5 and the rules are still work in progress.
232 * Many variables that points at a region of memory that will live
233 throughout the life of the program have been marked with UNLEAK
234 marker to help the leak checkers concentrate on real leaks..
236 * Plans for weaning us off of SHA-1 has been documented.
238 * A new "oidmap" API has been introduced and oidset API has been
239 rewritten to use it.
242 Also contains various documentation updates and code clean-ups.
245 Fixes since v2.14
246 -----------------
248 * "%C(color name)" in the pretty print format always produced ANSI
249 color escape codes, which was an early design mistake. They now
250 honor the configuration (e.g. "color.ui = never") and also tty-ness
251 of the output medium.
253 * The http.{sslkey,sslCert} configuration variables are to be
254 interpreted as a pathname that honors "~[username]/" prefix, but
255 weren't, which has been fixed.
257 * Numerous bugs in walking of reflogs via "log -g" and friends have
258 been fixed.
260 * "git commit" when seeing an totally empty message said "you did not
261 edit the message", which is clearly wrong. The message has been
262 corrected.
264 * When a directory is not readable, "gitweb" fails to build the
265 project list. Work this around by skipping such a directory.
267 * Some versions of GnuPG fails to kill gpg-agent it auto-spawned
268 and such a left-over agent can interfere with a test. Work it
269 around by attempting to kill one before starting a new test.
271 * A recently added test for the "credential-cache" helper revealed
272 that EOF detection done around the time the connection to the cache
273 daemon is torn down were flaky. This was fixed by reacting to
274 ECONNRESET and behaving as if we got an EOF.
276 * "git log --tag=no-such-tag" showed log starting from HEAD, which
277 has been fixed---it now shows nothing.
279 * The "tag.pager" configuration variable was useless for those who
280 actually create tag objects, as it interfered with the use of an
281 editor. A new mechanism has been introduced for commands to enable
282 pager depending on what operation is being carried out to fix this,
283 and then "git tag -l" is made to run pager by default.
285 * "git push --recurse-submodules $there HEAD:$target" was not
286 propagated down to the submodules, but now it is.
288 * Commands like "git rebase" accepted the --rerere-autoupdate option
289 from the command line, but did not always use it. This has been
290 fixed.
292 * "git clone --recurse-submodules --quiet" did not pass the quiet
293 option down to submodules.
295 * Test portability fix for OBSD.
297 * Portability fix for OBSD.
299 * "git am -s" has been taught that some input may end with a trailer
300 block that is not Signed-off-by: and it should refrain from adding
301 an extra blank line before adding a new sign-off in such a case.
303 * "git svn" used with "--localtime" option did not compute the tz
304 offset for the timestamp in question and instead always used the
305 current time, which has been corrected.
307 * Memory leak in an error codepath has been plugged.
309 * "git stash -u" used the contents of the committed version of the
310 ".gitignore" file to decide which paths are ignored, even when the
311 file has local changes. The command has been taught to instead use
312 the locally modified contents.
314 * bash 4.4 or newer gave a warning on NUL byte in command
315 substitution done in "git stash"; this has been squelched.
317 * "git grep -L" and "git grep --quiet -L" reported different exit
318 codes; this has been corrected.
320 * When handshake with a subprocess filter notices that the process
321 asked for an unknown capability, Git did not report what program
322 the offending subprocess was running. This has been corrected.
324 * "git apply" that is used as a better "patch -p1" failed to apply a
325 taken from a file with CRLF line endings to a file with CRLF line
326 endings. The root cause was because it misused convert_to_git()
327 that tried to do "safe-crlf" processing by looking at the index
328 entry at the same path, which is a nonsense---in that mode, "apply"
329 is not working on the data in (or derived from) the index at all.
330 This has been fixed.
332 * Killing "git merge --edit" before the editor returns control left
333 the repository in a state with MERGE_MSG but without MERGE_HEAD,
334 which incorrectly tells the subsequent "git commit" that there was
335 a squash merge in progress. This has been fixed.
337 * "git archive" did not work well with pathspecs and the
338 export-ignore attribute.
340 * In addition to "cc: <a@dd.re.ss> # cruft", "cc: a@dd.re.ss # cruft"
341 was taught to "git send-email" as a valid way to tell it that it
342 needs to also send a carbon copy to <a@dd.re.ss> in the trailer
343 section.
345 * "git branch -M a b" while on a branch that is completely unrelated
346 to either branch a or branch b misbehaved when multiple worktree
347 was in use. This has been fixed.
348 (merge 31824d180d nd/worktree-kill-parse-ref later to maint).
350 * "git gc" and friends when multiple worktrees are used off of a
351 single repository did not consider the index and per-worktree refs
352 of other worktrees as the root for reachability traversal, making
353 objects that are in use only in other worktrees to be subject to
354 garbage collection.
356 * A regression to "gitk --bisect" by a recent update has been fixed.
358 * "git -c submodule.recurse=yes pull" did not work as if the
359 "--recurse-submodules" option was given from the command line.
360 This has been corrected.
362 * Unlike "git commit-tree < file", "git commit-tree -F file" did not
363 pass the contents of the file verbatim and instead completed an
364 incomplete line at the end, if exists. The latter has been updated
365 to match the behaviour of the former.
367 * Many codepaths did not diagnose write failures correctly when disks
368 go full, due to their misuse of write_in_full() helper function,
369 which have been corrected.
370 (merge f48ecd38cb jk/write-in-full-fix later to maint).
372 * "git help co" now says "co is aliased to ...", not "git co is".
373 (merge b3a8076e0d ks/help-alias-label later to maint).
375 * "git archive", especially when used with pathspec, stored an empty
376 directory in its output, even though Git itself never does so.
377 This has been fixed.
379 * API error-proofing which happens to also squelch warnings from GCC.
381 * The explanation of the cut-line in the commit log editor has been
382 slightly tweaked.
383 (merge 8c4b1a3593 ks/commit-do-not-touch-cut-line later to maint).
385 * "git gc" tries to avoid running two instances at the same time by
386 reading and writing pid/host from and to a lock file; it used to
387 use an incorrect fscanf() format when reading, which has been
388 corrected.
390 * The scripts to drive TravisCI has been reorganized and then an
391 optimization to avoid spending cycles on a branch whose tip is
392 tagged has been implemented.
393 (merge 8376eb4a8f ls/travis-scriptify later to maint).
395 * The test linter has been taught that we do not like "echo -e".
397 * Code cmp.std.c nitpick.
399 * A regression fix for 2.11 that made the code to read the list of
400 alternate object stores overrun the end of the string.
401 (merge f0f7bebef7 jk/info-alternates-fix later to maint).
403 * "git describe --match" learned to take multiple patterns in v2.13
404 series, but the feature ignored the patterns after the first one
405 and did not work at all. This has been fixed.
407 * "git filter-branch" cannot reproduce a history with a tag without
408 the tagger field, which only ancient versions of Git allowed to be
409 created. This has been corrected.
410 (merge b2c1ca6b4b ic/fix-filter-branch-to-handle-tag-without-tagger later to maint).
412 * "git cat-file --textconv" started segfaulting recently, which
413 has been corrected.
415 * The built-in pattern to detect the "function header" for HTML did
416 not match <H1>..<H6> elements without any attributes, which has
417 been fixed.
419 * "git mailinfo" was loose in decoding quoted printable and produced
420 garbage when the two letters after the equal sign are not
421 hexadecimal. This has been fixed.
423 * The machinery to create xdelta used in pack files received the
424 sizes of the data in size_t, but lost the higher bits of them by
425 storing them in "unsigned int" during the computation, which is
426 fixed.
428 * The delta format used in the packfile cannot reference data at
429 offset larger than what can be expressed in 4-byte, but the
430 generator for the data failed to make sure the offset does not
431 overflow. This has been corrected.
433 * The documentation for '-X<option>' for merges was misleadingly
434 written to suggest that "-s theirs" exists, which is not the case.
436 * "git fast-export" with -M/-C option issued "copy" instruction on a
437 path that is simultaneously modified, which was incorrect.
438 (merge b3e8ca89cf jt/fast-export-copy-modify-fix later to maint).
440 * Many codepaths have been updated to squelch -Wsign-compare
441 warnings.
442 (merge 071bcaab64 rj/no-sign-compare later to maint).
444 * Memory leaks in various codepaths have been plugged.
445 (merge 4d01a7fa65 ma/leakplugs later to maint).
447 * Recent versions of "git rev-parse --parseopt" did not parse the
448 option specification that does not have the optional flags (*=?!)
449 correctly, which has been corrected.
450 (merge a6304fa4c2 bc/rev-parse-parseopt-fix later to maint).
452 * The checkpoint command "git fast-import" did not flush updates to
453 refs and marks unless at least one object was created since the
454 last checkpoint, which has been corrected, as these things can
455 happen without any new object getting created.
456 (merge 30e215a65c er/fast-import-dump-refs-on-checkpoint later to maint).
458 * Spell the name of our system as "Git" in the output from
459 request-pull script.
461 * Fixes for a handful memory access issues identified by valgrind.
463 * Backports a moral equivalent of 2015 fix to the poll() emulation
464 from the upstream gnulib to fix occasional breakages on HPE NonStop.
466 * Users with "color.ui = always" in their configuration were broken
467 by a recent change that made plumbing commands to pay attention to
468 them as the patch created internally by "git add -p" were colored
469 (heh) and made unusable. This has been fixed by reverting the
470 offending change.
472 * In the "--format=..." option of the "git for-each-ref" command (and
473 its friends, i.e. the listing mode of "git branch/tag"), "%(atom:)"
474 (e.g. "%(refname:)", "%(body:)" used to error out. Instead, treat
475 them as if the colon and an empty string that follows it were not
476 there.
478 * An ancient bug that made Git misbehave with creation/renaming of
479 refs has been fixed.
481 * Other minor doc, test and build updates and code cleanups.
482 (merge f094b89a4d ma/parse-maybe-bool later to maint).
483 (merge 6cdf8a7929 ma/ts-cleanups later to maint).
484 (merge 7560f547e6 ma/up-to-date later to maint).
485 (merge 0db3dc75f3 rs/apply-epoch later to maint).
486 (merge 276d0e35c0 ma/split-symref-update-fix later to maint).
487 (merge f777623514 ks/branch-tweak-error-message-for-extra-args later to maint).
488 (merge 33f3c683ec ks/verify-filename-non-option-error-message-tweak later to maint).
489 (merge 7cbbf9d6a2 ls/filter-process-delayed later to maint).